RIG-TARS open-source ESP32-S3 desktop robot ships $99 kit with local wake-word AI
LuwuDynamics' 3D-printable RIG-TARS robot runs Apache 2.0-licensed C++ firmware on the ESP32-S3, combining local wake-word detection with cloud speech services and Model Context Protocol tool support.
Summary
LuwuDynamics has released RIG-TARS, a 3D-printable desktop robot built around Espressif's ESP32-S3 microcontroller. The tripartite body walks and turns via three SCS009 serial servos controlled over UART, with position feedback from a six-axis QMI8658C IMU. Voice interaction runs through an onboard microphone and speaker: wake-word detection happens locally on the ESP32-S3, while speech recognition and synthesis are offloaded to cloud services. A 2.4-inch 320x240 color screen serves as the console, and the device supports Bluetooth LE control, over-the-air firmware updates, and Model Context Protocol tools.
The firmware, named RIG-Omni, is written in C++ on Espressif's ESP-IDF and is available on GitHub under the Apache 2.0 license. Notably, the same codebase targets four distinct hardware variants: the TARS walking robot, a robotic arm, a balancing robot, and a robot dog. The electronics kit retails for $99, with users responsible for 3D-printing the shell from mechanical design files included in the repository.
Hardware architecture and firmware design
- ESP32-S3 core: The robot uses Espressif's ESP32-S3 as its central microcontroller, running custom C++ firmware on the ESP-IDF framework.
- Actuation and sensing: Three SCS009 serial servos drive the three-segment body over UART, with a QMI8658C six-axis IMU providing motion and tilt feedback for closed-loop control.
- Local vs. cloud AI split: Wake-word detection runs on-device; speech recognition and text-to-speech synthesis are handled by cloud services, creating a hybrid edge-cloud architecture.
- Connectivity stack: The firmware supports Bluetooth LE for alternative control, over-the-air updates for maintenance, and implements Model Context Protocol for tool-use capabilities.
- Unified codebase: RIG-Omni targets four hardware form factors from a single repository: TARS walker, robotic arm, balancing robot, and quadruped dog.
Why it matters
The RIG-TARS release contributes a fully open-source mechanical and firmware stack to the desktop robotics space, lowering the barrier for hobbyists and educators to experiment with voice-controlled locomotion. Its Apache 2.0-licensed ESP-IDF codebase and shared multi-platform architecture could reduce fragmentation in small-scale robot firmware development, while the Model Context Protocol support positions it within emerging AI agent interoperability standards.
Original reference: blog.adafruit.com →